Money Laundering Activity Detection Method Based on Subgraph Pattern Search

Given a financial transaction network, how can we detect money laundering activities within it? Money laundering organizations often use decentralized, multi-layered small transactions to hide the source of their illicit funds and employ sophisticated strategies to evade detection. Existing money-laundering detection methods for transaction networks either ignore transaction attributes only by considering topological similarity, or can only identify fixed chain transactions based on K-part graph. Therefore, this paper proposes a money laundering detection method based on pattern subgraph search. By analyzing the flow of funds, this method uncovers money laundering subgraphs hidden within the transaction flows. We model money laundering activities as pattern graphs and design a new scoring function to measure the degree to which the given subgraph conforms to the characteristics of money laundering. Experimental results show that this method outperforms decision trees, random forests, and naive Bayes in terms of precision, recall, and F1 score.